Experimental Design 2

define experimental design

the process of carrying out research in an objective and controlled fashion so precision is maximised and specific conclusions can be drawn about a pre-existing hypothesis

good design = valid conclusion

research cycle

what is the aim of a ‘true experiment’?

to establish the effect of an IV on an outcome measure, with only one difference between groups and everything else remaining constant to minimise variation

define hypothesisb

a testable prediction, likely based on literature or a previous experiment 

what makes a good hypothysis?

• states what you expect to find, with a direction

• be testable

• contain an IV and an outcome measure (DV)

7 key choices to make in experimental design

• subjects

• techniques

• control groups

• randomisation

• sample size

• statistical analysis

advantages of cells as a subject

• give insight into molecular mechanisms, such as mitochondrial function and genetics 

disadvantages of cells as a subject

lack of peripheral influence (eg. from gut microbiome-brain axis and hypothalamic-pituitary-adrenal axis) 

advantages of animals as subjects

use of whole animals to do invasive studies which can demonstrate cause and effect

disadvantages of animals as subjects

• ethical considerations

• differences between species, so a lack of translation to humans

  • vertebrates and invertebrates differ, so mammals may be required

  • need correct brain region, neurological chemicals or behaviours

advantages of humans as subjects 

relevance to complex human emotion and disease

disadvantages of humans as subjects

• regulatory constraints

• recruitment difficulties

• difficulty controlling all variables

factors to consider when choosing subjects

• ethics

• time

• finances

choosing species in animal studies

• the legal and ethical obligation is to use the simplest and least sentient species appropriate for the scientific objective

• must have the correct brain regions, chemicals and behaviours

why are primates used to study the granular frontal cortex

• rodents lack a granular frontal cortex

• this is involved in higher cognitive functions (conceptualising, planning and prioritising)

• these behaviours cannot be studied in rodents

• this justifies the use of primates

choosing strains for rodent studies

• there is no standard lab rat or mouse

• different strains differ in sight, activity, sociability, stress response and cognition

• therefore different strains are preferred for different types of study

choosing the age of subjects in an animal experiment

• lifespan and developmental milestones differ based on species

• in humans, apoptosis and myelination begin in utero whereas this starts postnatally in rodents 

• therefore experimental interventions and data collection need to be timed appropriately to be relevant 

effects of supplier and in-house conditions when selecting rodents for a study

• rodents can be kept differently, husbandry differs by institution

• this can affect the outcome 

  • more enriched cages can encourage more neurogenesis

• this can cause different baseline conditions

choosing the sex of animal subjects 

• females are no longer viewed as more variable

• both make and female are starting to be used unless there is a specific reason not to

mouse handling methods

• must be used to handling before the experiment

• picking up by the tail causes anxiety and aversion so the mouse will not engage

  • reduces willingness to explore in behavioural testing

  • impacts data

• Picking up by tunnel or cup is preferred as allows habituation

role of studies using normal animals

• illustrate physiological roles of neurotransmitters

• cannot predict efficacy 

role of studies using disease models

• use genetic, environmental or surgical manipulation to mirror known risks and symptoms

• provide better predictions of clinical efficacy 

choosing behavioural tests 

• there are multiple tests for different things like pain, cognition etc

• the cognitive domain and disease being tested changes the relevant testing technique 

• some domains cannot be tested in animals 

• different cognitive tests fit different diseases better

Wisconsin card sorting test

• for humans

• test executive function

• cards are sorted by a category, which changes

• time for new rule to be realised

• relevant in schizophrenia, ASD and ADHD

• less relevant in PTSD and AD

Attentional set shifting task test

• used in rodents 

• tests executive function

• rule for what food us under (material and smell) is changed 

• time for new rule to be realised 

• relevant in schizophrenia, ASD and ADHD

• less relevant in PTSD and AD

Control group

• everything is the same as it is for the experimental group, but the IV is not manipulated 

why are positive and negative controls necessary?

• show that the assay correctly detects the outcome measure correctly

• do not actually address the hypothesis

Positive control

• shows that the experiment could detect a positive effect if it was to occur

• is reassuring if the IV has no effect

Negative control

• shows that the experiment does not give artificial positive effects 

• is reassuring if the IV increases readout 

Sample size definition

the number of experimental units per group

Experimental Unit

the entity subjected to an intervention independently of all other units

• eg. number of animals, humans or brain slices

• offspring are one unit if the intervention was done on the parent

• one cell stock is one unit

Deciding how many experimental units are needed

• dependent on the expected data → could be calculated from a prior experiment

• group needs to be large enough to produce biologically relevant data between groups

• predict variability between groups

• depends on experimental parameters

  • significance level (usually 0.05 → <5% chance of detecting false effect)

  • power (usually 0.8 to 0.95 → 80-95% chance of detecting a real effect) 

    • calculated using NC3Rs Experimental Design Assistant (EDA)

What are the 3Rs

• Replacement 

  • avoid or replace use of animals

• Reduction

  • minimise number of animals per experiment

• Refinement

  • minimise animal suffering and improve welfare

Reduction of animals used in experiments

• need proper sample size → without proper sample calculations results could be impossible to interpret, wasting all the animals

• sample size calculations allow robust results

• report in detail so people can follow methods without wasting animals

Randomisation

• groups must be equivalent except for the IV

• experiments must avoid introducing confounds → randomisation does this 

Circadian variations in behaviour

• rodents are nocturnal

  • more active in dark conditions

  • randomise testing times for mice so that this doesnt cause a confound

why is blinding important?

• expecting a particular outcome makes researchers unintentionally influence data collection or their interpretation of data 

• this is subconscious bias

Open label human trial

both the researchers and participants know the treatment

double-blind human trial

• neither the researcher or participants know the treatment

• this minimises bias

perceived barriers

things that effect the accuracy and validity of research 

3 methods of blinding

• Partial binding, single handler

  • person A prepares the drug, person B doses and tests animals

• multiple handlers, full blinding

  • person A prepares drugs and doses animals, person B tests animals

• Automated data collection → desirable as animal appearance doesn’t prevent blinding

  • must be declared

Deciding statistical tests 

• should be decided before the experiment starts to spot and avoid design flaws → can’t collect data that can’t be analysed or interpreted